Tuesday, December 08, 2009
Derek Thompson at The Atlantic has a post Are America's Fattest States Also the Most Jobless?. The county-level data on unemployment only goes back to 2008 (at least that I can find online). But I do have data on obesity at the county-level too. What's the correlation? 0.32. Pretty modest. If I correlate for white obesity it goes down a little, 0.23 (though remember that I estimated white obesity, so be cautious about this). Since I also have food stamp utilization data I looked at that. Correlation is 0.56. If you think of this as r-squared, how much of variance of Y can be explained by X by squaring the correlation, it's a much stronger association. I constructed a quick regression where % unemployed on the county-level was the dependent variable, and % black, obese, median household income and % on food stamps were the independents. Except for food stamps none of these variables generated statistically significant beta coefficients. In other words, regional level differences in unemployment in 2008 which tracked obesity are probably best explained as emerging out of a general poverty factor (though do note that median household income itself isn't very predictive once % on food stamps gets put into the equation).
I don't doubt that all things equal the obese would be fired first. That being said, all things are often not equal.
Update: I realized I left something out. Looking at the correlation college degree holding on the county-level and unemployment in 2008, I found it to be -0.43. So I popped that into the regression, and here are the coefficients with standard errors (all statistically significant):
Black 1.20987610 (0.33416977)
College Degree -7.64273043 (0.62394667)
Percent on Food Stamps 0.14095962 (0.00946762)
Median Household Income 0.00002967 (0.00000523)
Obesity -0.06840311 (0.01494881)
I'll let readers wonder what's going on here, though I assume it has something to do with the changes in the education premium and such with globalization.